Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A system, comprising: one or more computing devices of a network-accessible computing service; wherein the one or more computing devices comprise program instructions that when executed on or across one or more processors cause the one or more computing devices to: obtain respective indications of (a) a machine image to be used to instantiate one or more micro virtual machines which meet a first resource capacity criterion, (b) an isolated virtual network to which connectivity from the one or more micro virtual machines is to be established, and (c) a first virtualization control mode associated with the one or more micro virtual machines; initiate, based at least in part on the first virtualization control mode, one or more preparatory configuration operations enabling an accelerated response to a future launch request for a micro virtual machine, wherein the one or more preparatory configuration operations include (a) a transfer of the machine image to a storage device accessible from a first virtualization host and (b) an initialization of one or more virtual network interfaces at the first virtualization host, including a first virtual network interface; and cause, responsive to request and subsequent to initiating the one or more preparatory configuration operations, a first micro virtual machine to be instantiated using the machine image at the first virtualization host, wherein the first micro virtual machine is a thread of an intermediary process launched by an administrative agent of the computing service at the first virtualization host in response to a first launch request, and wherein the first micro virtual machine is configured within the isolated virtual network.
2. The system as recited in claim 1 , wherein a second launch request indicates that a plurality of collaborating micro virtual machines, including a second micro virtual machine and a third micro virtual machine, are to be instantiated, wherein the one or more computing devices further comprise program instructions that when executed on or across one or more processors further cause the one or more computing devices to: cause the second and third micro virtual machines to be launched at the first virtualization host; and causing a local channel for communication between the second and third virtual machines to be established at the first virtualization host, wherein the local communication channel does not utilize a network device.
3. The system as recited in claim 1 , wherein the one or more computing devices further comprise program instructions that when executed on or across one or more processors further cause the one or more computing devices to: provide, based at least in part on determining that the first virtualization control mode is in use, respective identifiers of one or more candidate virtualization hosts to a first client of a computing service, wherein the first launch request comprises an identifier of the first virtualization host, and wherein an identifier of a candidate virtualization host for a micro virtual machine to be configured using a second virtualization control mode on behalf of a second client is not provided to the second client.
4. The system as recited in claim 1 , wherein the one or more computing devices further comprise program instructions that when executed on or across one or more processors further cause the one or more computing devices to: obtain an indication, via a programmatic request from a client, of an identifier of a virtualization host to be used for a second virtual machine, wherein instantiation of the second virtual machine results in an oversubscription of at least one type of resource.
5. The system as recited in claim 1 , wherein the one or more computing devices further comprise program instructions that when executed on or across one or more processors further cause the one or more computing devices to: based at least in part on determining that the first virtualization control mode is to be used for the one or more micro virtual machines, store metadata indicating that one or more candidate virtualization hosts are dedicated exclusively for a client.
6. A method, comprising: performing, by one or more computing devices: determining a first virtualization control mode to be used for configuring one or more compute instances; initiating, responsive to determining the first virtualization control mode, one or more preparatory configuration operations enabling an accelerated response to a future request for launching at least one compute instance of the one or more compute instances, including (a) a transfer of at least a portion of a first machine image to a storage device accessible from a first virtualization host selected according to the determined first virtualization control mode, and (b) an initialization of one or more virtual network interfaces; and causing, responsive to request and subsequent to initiating the one or more preparatory configuration operations, a first compute instance to be launched using the first machine image at the first virtualization host, wherein the first compute instance is a thread of an intermediary process launched by an administrative agent at the first virtualization host in response to a first launch request.
7. The method as recited in claim 6 , wherein determining the first virtualization control mode comprises obtaining an indication of the first virtualization control mode via a programmatic interface.
8. The method as recited in claim 6 , further comprising performing, by the one or more computing devices: providing, based at least in part on the first virtualization control mode, respective identifiers of one or more candidate virtualization hosts to a first client of a computing service, wherein the first launch request comprises an identifier of the first virtualization host, and wherein an identifier of a candidate virtualization host for a compute instance to be configured using a second virtualization control mode on behalf of a second client is not provided to the second client.
9. The method as recited in claim 6 , wherein the first virtualization host has a first resource capacity of a first resource type, the method further comprising performing, by the one or more computing devices: providing, to a client of a computing service on whose behalf the first compute instance is launched, respective indicators of (a) requested resource levels, with respect to the first resource type, of a set of compute instances of the first virtualization host, and (b) a metric of measured resource usage, with respect to the first resource type, of the set of compute instances; and causing, in response to an additional launch request for an additional compute instance, the additional compute instance to be instantiated at the first virtualization host, without verifying whether a requested resource level of the additional compute instance with respect to the first resource type would cause the first resource capacity to be exceeded.
This invention relates to resource management in virtualized computing environments, specifically addressing the challenge of efficiently allocating and monitoring compute instance resources without strict pre-launch capacity checks. The system involves a first virtualization host with a defined resource capacity of a particular resource type (e.g., CPU, memory). The method includes providing a client of a computing service with indicators of both the requested resource levels for a set of compute instances hosted on the first virtualization host and a metric of measured resource usage for those instances. When an additional compute instance is requested, the system instantiates it on the first virtualization host without verifying whether the requested resource level for the additional instance would exceed the host's capacity. This approach allows for dynamic resource allocation, potentially enabling overcommitment of resources while relying on runtime monitoring to manage performance. The system may also involve other virtualization hosts and compute instances, with similar resource tracking and allocation mechanisms. The invention aims to optimize resource utilization and responsiveness in cloud computing environments by reducing pre-launch validation overhead.
10. The method as recited in claim 9 , wherein the first resource type comprises one or more of: (a) processors, (b) volatile or non-volatile memory, (c) persistent storage, or (d) network bandwidth.
This invention relates to resource management in computing systems. It addresses the problem of efficiently allocating and controlling different types of computing resources. The method involves a system that manages computing resources. Specifically, it differentiates between at least two types of resources: a first resource type and a second resource type. The system operates such that when a particular allocation or operation is performed concerning the first resource type, this action triggers a corresponding operation related to the second resource type. The first resource type can include one or more of processors, volatile memory, non-volatile memory, persistent storage, or network bandwidth. The core of the invention lies in this cross-resource dependency, where managing one category of resource directly influences the handling of another distinct category.
11. The method as recited in claim 6 , further comprising performing, by the one or more computing devices: storing metadata indicating that one or more candidate virtualization hosts are designated exclusively for compute instances requested by a first client of a computing service.
12. The method as recited in claim 11 , further comprising performing, by the one or more computing devices: identifying, in response to a second launch request, a particular virtualization host of the one or more candidate virtualization hosts at which a second compute instance is to be instantiated, wherein the second launch request does not include an indication of a target virtualization host to be used for the second compute instance.
This invention relates to virtualization systems, specifically methods for dynamically selecting virtualization hosts for compute instance instantiation. The problem addressed is the inefficiency in traditional systems where compute instances are launched on predetermined hosts, leading to suboptimal resource utilization and potential performance bottlenecks. The method involves a system with multiple candidate virtualization hosts, each capable of running compute instances. When a launch request is received for a new compute instance, the system evaluates the current state of the candidate hosts, including their resource availability, performance metrics, and workload characteristics. Based on this evaluation, the system selects an optimal host for instantiating the compute instance, even when the launch request does not specify a target host. This dynamic selection ensures that compute instances are placed on hosts that can provide the best performance and resource efficiency. The system also includes mechanisms for monitoring the performance of instantiated compute instances and adjusting their placement if necessary. If a compute instance is underperforming or if resource conditions change, the system can migrate the instance to a different host to improve efficiency. Additionally, the system can prioritize hosts based on factors such as energy efficiency, hardware capabilities, or network proximity to other instances, further optimizing resource allocation. This approach improves overall system efficiency, reduces manual intervention, and enhances the scalability of virtualized environments.
13. The method as recited in claim 11 , further comprising performing, by the one or more computing devices: based at least in part on determining that a second virtualization control mode is to be used for a second compute instance on behalf of a second client, causing a second compute instance, requested by the second client, to be instantiated at a particular virtualization host; and causing a third compute instance, requested by the third client, to be instantiated at the particular virtualization host.
14. The method as recited in claim 6 , wherein the first launch request indicates that a plurality of collaborating compute instances, including the first compute instance and a second compute instance, is to be instantiated, the method further comprising performing, by the one or more computing devices: causing the second compute instance to be launched at the first virtualization host; and causing a local channel for communication between the first and second compute instances to be established at the first virtualization host, wherein the local communication channel does not utilize a network device.
15. The method as recited in claim 6 , further comprising performing, by the one or more computing devices: pausing execution of the first compute instance; saving, to a persistent storage device, a snapshot of state information of the first compute instance; and in response to a programmatic request, launching a clone compute instance of the first compute instance.
16. One or more non-transitory computer-accessible storage media storing program instructions that when executed on one or more processors cause one or more computer systems to: determine a first virtualization control mode to be used for configuring one or more compute instances; initiate, based at least in part on the first virtualization control mode, one or more preparatory configuration operations enabling an accelerated response to a future request for launching at least one compute instance of the one or more compute instances, including (a) a transfer of at least a portion of a first machine image to a storage device accessible from a first virtualization host selected based at least in part on the first virtualization control mode, and (b) an initialization of one or more virtual network interfaces to be used at the first virtualization host to communicate with one or more endpoints; and cause, responsive to request and subsequent to initiating the one or more preparatory configuration operations, a first compute instance to be launched using the first machine image at the first virtualization host, wherein the first compute instance is a thread of an intermediary process launched by an administrative agent at the first virtualization host in response to a first launch request.
17. The one or more non-transitory computer-accessible storage media as recited in claim 16 , further storing program instructions that when executed on one or more processors further cause the one or more computer systems to: obtain an indication of the first virtualization control mode via a programmatic interface.
18. The one or more non-transitory computer-accessible storage media as recited in claim 16 , wherein the first machine image is generated by a machine image optimization tool of a computing service, wherein to generate the first machine image, the machine image optimization tool eliminates, from a baseline image, based at least in part on a usage analysis, one or more of: (a) a kernel module, or (b) a system library.
19. The one or more non-transitory computer-accessible storage media as recited in claim 16 , further storing program instructions that when executed on one or more processors further cause the one or more computer systems to: in response to a mode change request submitted via a programmatic interface, store metadata indicating that a second virtualization control mode is to be used at the first virtualization host; and cause, in accordance with the second virtualization control mode, an additional compute instance to be instantiated at the first virtualization host.
20. The one or more non-transitory computer-accessible storage media as recited in claim 16 , further storing program instructions that when executed on one or more processors further cause the one or more computer systems to: provide an indication, via a programmatic interface, of a lifetime limit of one or more compute instances including the first compute instance; and cause the first compute instance to be automatically terminated in response to a determination that the lifetime limit has been reached by the first compute instance.
Unknown
March 30, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.